Skip to content
LumenaryAcademy
Frontier 85 min

Sustainable chemistry

Capturing and converting CO₂, green hydrogen and ammonia, chemical recycling of plastics, replacing PFAS, critical elements, and life-cycle assessment — the chemistry, the energy bills and the honest trade-offs as of late 2026.

Code tested · 6AI-draftedAwaiting review by a chemist

Your guides: Chloe · The Chemist, Margaret · The Analyst, Gideon · The Sceptic

MotivationH

The second law sends the bill

Catalysis and green chemistry gave the principles and metrics a chemist uses at the bench: prevent waste, maximise atom economy, choose safer solvents, think in life cycles. This lesson is about the same ideas at the scale of the planet, where the questions are measured in gigatonnes and exajoules.

The scale is the first fact. The chemical industry is the largest industrial consumer of energy and among the largest industrial emitters of CO₂ (IEA). The world made about 460 million tonnes of plastics in 2019, and only 9 per cent of plastic waste was recycled (OECD). It makes close to 100 million tonnes of hydrogen a year, almost all from fossil fuels (IEA), much of it for ammonia.

The second fact is thermodynamic. Most of the hard problems here are problems of un-mixing: pulling CO₂ out of air at 0.04 per cent, sorting a tangle of polymers back into pure monomers, recovering a few grams of rare earth from a tonne of scrap, or gathering persistent molecules from a river. Un-mixing lowers entropy, so it costs work, and Entropy and free energy tells us exactly how much. The rest is engineering, economics and honesty about trade-offs. This is a briefing as of late 2026, written to inform: the numbers matter more than the slogans.

DefinitionsH

Capturing carbon dioxide

Definition— Carbon capture, removal and storage

Point-source capture separates CO₂ from a concentrated stream — the flue gas of a power station or cement kiln, or a natural-gas or hydrogen plant — before it reaches the air. Direct air capture (DAC) separates it from ambient air. Captured CO₂ may be stored, in deep saline rock formations or by mineralisation, or used. Only capture from air or from biomass followed by durable storage is carbon dioxide removal: it lowers the amount in the atmosphere rather than slowing its rise.

Theorem— Minimum work of separation

At constant temperature and pressure, the least work needed to separate an ideal-gas mixture into product streams is Wmin⁡=RT[∑products∑iniln⁡yi−∑iniln⁡xi]W_{\min} = RT\left[\sum_{\text{products}} \sum_i n_i \ln y_i - \sum_i n_i \ln x_i\right], where xix_i are mole fractions in the feed and yiy_i in each product. Removing a small amount of a component present at mole fraction xx, as the pure gas, costs −RTln⁡x-RT\ln x per mole.

Proof

The minimum work at constant TT and pp is the increase in Gibbs energy. For an ideal gas μi=μi∘+RTln⁡yi\mu_i = \mu_i^\circ + RT\ln y_i, so a stream's Gibbs energy is ∑iniμi∘+RT∑iniln⁡yi\sum_i n_i\mu_i^\circ + RT\sum_i n_i \ln y_i; the μi∘\mu_i^\circ terms are the same before and after and cancel, which gives the formula. Moving δn\delta n from a large mixture at μ∘+RTln⁡x\mu^\circ + RT\ln x into a pure stream at μ∘\mu^\circ raises GG by −RTln⁡x δn-RT\ln x\,\delta n.

∎

Amine scrubbing

The workhorse of point-source capture, used since the 1930s to clean natural gas, is a loop of aqueous amine (Rochelle, 2009). In an absorber at about 40 °C, a primary amine such as 2-aminoethan-1-ol (monoethanolamine, MEA) reacts with CO₂ to form a carbamate: 2 RNH₂ + CO₂ ⇌ RNHCOO⁻ + RNH₃⁺. The reaction is exothermic, so Le Chatelier's principle (Chemical equilibrium) says heat will reverse it: in a stripper at about 120 °C, steam from a reboiler drives the CO₂ back out, and the regenerated amine returns to the absorber.

absorberabout 40 °Cstripperabout 120 °Ccross heatexchangerreboilerflue gastreated gasrich amine (carbamate)lean amineCO₂ to compressionsteam
An amine scrubbing loop. Flue gas meets lean amine in the absorber, where CO₂ is taken up as carbamate. The rich amine is warmed by the hot lean amine in a cross heat exchanger, then stripped of CO₂ with steam from a reboiler; the lean amine returns to the absorber. Nearly all the energy cost is the reboiler steam.

The program computes the thermodynamic minimum for three streams, a rough reboiler duty for 30 wt % MEA, and a published direct-air-capture design.

python
import numpy as np

R = 8.314                 # J/(mol K)
M_CO2 = 44.01e-3          # kg/mol

def mix_term(amounts):
    """Sum of n_i ln y_i for one ideal-gas stream (amounts in mol); zero for a pure stream."""
    n = np.array([a for a in amounts if a > 0], float)
    return float(np.sum(n * np.log(n / n.sum())))

def w_min(x_co2, capture, T=298.15):
    """Minimum work (kJ per mol CO2 captured) to take a fraction `capture` of the CO2
    out of an ideal-gas mixture with CO2 mole fraction x_co2, as pure CO2, at constant T and p."""
    feed = [x_co2, 1 - x_co2]                                  # per mol of feed
    left = [x_co2 * (1 - capture), 1 - x_co2]                  # what goes back out
    captured = x_co2 * capture
    dG = R * T * (mix_term(left) + mix_term([captured]) - mix_term(feed))
    return dG / captured / 1000

print("minimum work of separation, pure CO2 product, 25 °C")
for name, x, phi in [("coal flue gas", 0.12, 0.90), ("gas-turbine flue gas", 0.04, 0.90),
                     ("air, 420 ppm, take half", 420e-6, 0.50), ("air, 420 ppm, take a little", 420e-6, 1e-6)]:
    w = w_min(x, phi)
    print(f"  {name:28s} {w:5.1f} kJ/mol = {w / M_CO2 / 1e3:.2f} GJ/t")

# A rough reboiler duty for 30 wt% aqueous MEA (2-aminoethan-1-ol), per mol of CO2 released.
mea = 300 / 61.08                     # mol MEA per kg of solution
cycled = 0.25 * mea                   # mol CO2 moved per kg of solution (loading swing 0.25 mol/mol)
sensible = 3.6 * 10 / cycled          # kJ: heat 1 kg (cp 3.6 kJ/(kg K)) by the last 10 K
desorb = 85.0                         # kJ/mol: reverses the exothermic absorption
steam = 0.8 * 40.7                    # kJ: water vapour that leaves the stripper with each mol of CO2
total = sensible + desorb + steam
print(f"\nMEA reboiler: sensible {sensible:.0f} + desorption {desorb:.0f} + stripping steam {steam:.0f}"
      f" = {total:.0f} kJ/mol = {total / M_CO2 / 1e3:.2f} GJ/t")
carnot = 1 - 298.15 / 393.15          # the most work that heat at 120 °C could have given
print(f"as work: x {carnot:.2f} (Carnot, 120 °C to 25 °C) = {total * carnot / M_CO2 / 1e3:.2f} GJ/t;"
      f" the minimum is {w_min(0.12, 0.90) / M_CO2 / 1e3:.2f} GJ/t")

# Direct air capture: a published design (Keith et al., 2018) against the minimum.
dac = 5.25 + 366 * 3.6e-3             # GJ/t: natural gas plus electricity, CO2 delivered at 15 MPa
compress = R * 298.15 * np.log(150) / M_CO2 / 1e6   # GJ/t, isothermal ideal compression 1 -> 150 bar
floor = w_min(420e-6, 0.5) / M_CO2 / 1e3 + compress
print(f"\nDAC design: {dac:.2f} GJ/t; minimum incl. compression to 15 MPa: {floor:.2f} GJ/t;"
      f" ratio {dac / floor:.0f}")
print(f"1 Gt CO2 a year at {dac:.1f} GJ/t = {dac:.1f} EJ/yr, {dac / 600:.1%} of ~600 EJ world primary energy")
Runs entirely in your browser, in an isolated worker. Nothing is sent to a server.

The minimum for coal flue gas is 0.16 GJ/t; for air, about 0.45 GJ/t — less than three times more, as the logarithm promised. A 30 wt % MEA plant needs about 3.3 GJ of heat per tonne in this rough model, close to the 3.5 to 4 GJ/t reported for first-generation plants; concentrated piperazine and other advanced solvents reach about 2.5. Heat is not work, though: at 120 °C its Carnot-limited work value is 0.81 GJ/t, so the real process is about five times the minimum.

In the IEA's tracking, about 27 DAC plants had been commissioned worldwide by the mid-2020s, capturing together on the order of 0.01 million tonnes of CO₂ a year, while all point-source capture with storage was just over 50 million tonnes a year in early 2025 (IEA). The approaches include liquid hydroxide solvents, solid amine sorbents regenerated by heat and vacuum near 100 °C, and framework sorbents such as the COFs of New reactions and framework materials. Storage is the better-proven half: Norway's Sleipner project has injected CO₂ into a saline formation under the North Sea since 1996, and in Iceland the CarbFix project dissolves CO₂ in water and injects it into basalt, where over 95 per cent mineralised within two years.

Conversion. CO₂ is a low point of chemical energy, so turning it into fuels needs at least as much energy as the fuel later releases. Its largest use is already old: urea, from ammonia and CO₂. Methanol from CO₂ and hydrogen is made commercially in Iceland and China, and electrochemical reduction to CO, methanoate (formate) or ethene on copper is an intense research area whose obstacles are selectivity, durability and the loss of CO₂ as carbonate in alkaline cells. A fuel made from captured CO₂ releases it again when burned; only durable products, such as carbonates in concrete, store carbon.

ComputationC

Green hydrogen and ammonia

Electrochemistry showed that splitting water needs 39.4 kWh per kilogram of hydrogen at the thermoneutral voltage and about 50 kWh/kg in real stacks. Four technologies compete in 2026. Alkaline electrolysers (nickel electrodes, potassium hydroxide solution) are mature and cheap. Proton-exchange-membrane (PEM) electrolysers run at high current density and follow a fluctuating supply well, but their acidic anode needs iridium oxide and their membrane is a perfluorosulfonic acid (PFSA) polymer. Solid-oxide cells run at 700 to 850 °C and need less electricity if heat is available; anion-exchange-membrane cells aim for PEM's performance without the platinum-group metals.

power supplye⁻e⁻water inO₂ outH⁺H₂ outanodeIrO₂ catalystmembranePFSA fluoropolymercathodePt catalystanode: 2 H₂O → O₂ + 4 H⁺ + 4 e⁻cathode: 4 H⁺ + 4 e⁻ → 2 H₂
A proton-exchange-membrane electrolyser. Water is oxidised at the iridium oxide anode to oxygen and protons; the protons cross the fluoropolymer membrane and are reduced to hydrogen at the platinum cathode, using the electrons that the power supply pushes round the external circuit.

The program follows the electricity from hydrogen to ammonia and methanol, and asks how much iridium a PEM build-out would need.

python
F = 96485.332                     # C/mol
M_H2, M_NH3 = 2.016, 17.031       # g/mol
dH = 285.83                       # kJ/mol, H2O(l) -> H2(g) + 1/2 O2(g): the higher heating value of H2

def kwh_per_kg(V):
    """Electrical energy per kg of H2 for a cell at voltage V (100 % Faradaic efficiency)."""
    return 2 * F * V / (M_H2 / 1000) / 3.6e6

E_tn = dH * 1e3 / (2 * F)         # thermoneutral voltage, 1.481 V
for V in [1.8, 2.0]:
    print(f"stack at {V} V: {kwh_per_kg(V):4.1f} kWh/kg H2, efficiency (HHV) {E_tn / V:.0%}")

# Green ammonia: N2 + 3 H2 -> 2 NH3, so 1.5 mol H2 per mol NH3.
system = 52.0                                          # kWh/kg H2: an assumed whole-plant electrolyser figure
h2_per_t = 1e6 * 1.5 * M_H2 / M_NH3 / 1000            # kg H2 per tonne NH3
mwh = h2_per_t * system / 1000 + 0.5                   # + an assumed 0.5 MWh/t for air separation and the loop
print(f"\nammonia: {h2_per_t:.1f} kg H2 per tonne; about {mwh:.1f} MWh of electricity per tonne")
print(f"all ~185 Mt/yr of ammonia this way: {185e6 * mwh / 1e6:,.0f} TWh,"
      f" {185e6 * mwh / 1e6 / 30000:.0%} of ~30,000 TWh world electricity")

# Power-to-methanol: CO2 + 3 H2 -> CH3OH + H2O. Energy out (HHV) against energy in.
dcH_meoh = 726.0                                       # kJ/mol, enthalpy of combustion of liquid methanol
elec = 3 * M_H2 / 1000 * system * 3600                 # kJ of electricity for 3 mol H2
dac = 6.57e6 * 44.01e-6                                # kJ to capture 44 g of CO2 from air at 6.57 GJ/t
print(f"\nmethanol keeps {dcH_meoh / (3 * dH):.0%} of its hydrogen's energy;"
      f" from electricity and air, {dcH_meoh / (elec + dac):.0%} of the energy put in")

# Iridium for proton-exchange-membrane (PEM) electrolysers.
supply = 7.5e6                                         # g of iridium mined per year, roughly
for g_per_kw in [0.4, 0.1]:
    print(f"at {g_per_kw} g Ir/kW, a whole year's iridium builds {supply / g_per_kw / 1e6:4.1f} GW")
gw = 95e9 * system / (8760 * 0.5) / 1e6                # GW to replace 95 Mt/yr at a 50 % capacity factor
print(f"replacing ~95 Mt/yr of fossil hydrogen, running half the time: about {gw:,.0f} GW of electrolysers")
Runs entirely in your browser, in an isolated worker. Nothing is sent to a server.

Making all of today's ammonia from electrolytic hydrogen would take about 1,800 TWh a year, some 6 per cent of the world's electricity. Methanol made from air and electricity returns about half of the energy put in, which is why such fuels make sense mainly where batteries and direct electrification cannot reach, such as long-distance shipping and aviation. And iridium is a real constraint: world supply is roughly 7.5 tonnes a year, and at today's loading of about 0.4 g/kW a whole year's supply builds under 19 GW, against the thousand-plus gigawatts that replacing fossil hydrogen would need. Thinner catalyst layers, recycling and alkaline or anion-exchange designs, which need no iridium, are the responses (Johnson Matthey gives an industry view).

Ammonia. Green ammonia is the Haber–Bosch loop of Catalysis and green chemistry fed with electrolytic hydrogen; the first plants of hundreds of megawatts were being commissioned in 2025–26. The harder dream is to reduce N₂ electrochemically at room temperature. Early reports of this were often artefacts — ammonia and nitrogen oxides are everywhere in laboratories, and the signals were tiny — until a rigorous protocol with ¹⁵N₂ isotope controls showed which claims survived. Lithium-mediated nitrogen reduction, including a 2021 system with a phosphonium proton shuttle, is the credible route so far, still far from industrial scale.

Worked examplesH

Chemical recycling of plastics

Mechanical recycling — sort, wash, shred, melt, re-form — is the cheapest and usually the lowest-carbon route, but each melt cuts chains, additives and colours accumulate, and mixed or multilayer packaging cannot be sorted at all. Chemical recycling is a family of processes that change the polymer's chemistry: depolymerisation back to monomer, solvolysis of condensation polymers such as PET by water, methanol or glycols, purification by dissolving and reprecipitating the polymer, and pyrolysis or gasification of mixed waste to oils or synthesis gas.

Whether a polymer can return to its monomer is a question of equilibrium. Addition polymerisation is exothermic (ΔHp<0\Delta H_p \lt 0) and lowers entropy (ΔSp<0\Delta S_p \lt 0), so ΔGp=ΔHp−TΔSp\Delta G_p = \Delta H_p - T\Delta S_p changes sign at the ceiling temperature Tc=ΔHp/ΔSpT_c = \Delta H_p/\Delta S_p. Above it, the monomer is favoured; more precisely, the equilibrium monomer concentration is [M]eq=exp⁡(ΔHp/RT−ΔSp∘/R)[\text{M}]_{\text{eq}} = \exp(\Delta H_p/RT - \Delta S_p^\circ/R).

python
import numpy as np

R = 8.314
# Representative enthalpies (kJ/mol) and entropies (J/(mol K)) of polymerisation, monomer at 1 mol/L
# to polymer. Literature values differ by about 10 % with the states chosen.
polymers = {
    "poly(methyl methacrylate)": (-56.0, -117.0),
    "polystyrene":               (-73.0, -104.0),
    "poly(alpha-methylstyrene)": (-35.0, -104.0),
}

def ceiling(dH, dS):
    """Ceiling temperature (K) at 1 mol/L of monomer, where dG = dH - T dS = 0."""
    return dH * 1e3 / dS

def monomer_eq(dH, dS, T):
    """Equilibrium monomer concentration (mol/L) above the polymer: ln[M] = dH/RT - dS/R."""
    return np.exp(dH * 1e3 / (R * T) - dS / R)

print("equilibrium monomer concentration [M]eq in mol/L")
print(f"{'polymer':27s} {'Tc/°C':>6s} {'25 °C':>9s} {'150 °C':>9s} {'300 °C':>9s}")
for name, (dH, dS) in polymers.items():
    eq = [monomer_eq(dH, dS, T + 273.15) for T in (25, 150, 300)]
    print(f"{name:27s} {ceiling(dH, dS) - 273.15:6.0f} {eq[0]:9.1e} {eq[1]:9.1e} {eq[2]:9.1e}")
Runs entirely in your browser, in an isolated worker. Nothing is sent to a server.

Poly(methyl methacrylate) (acrylic sheet) has a ceiling near 205 °C; at 300 °C its equilibrium monomer concentration, about 10 mol/L, exceeds that of the pure liquid monomer (about 9.4 mol/L), so the polymer cannot survive: heated, it "unzips" to methyl methacrylate in high yield, and commercial plants do exactly this. Polystyrene's ceiling is much higher and its pyrolysis gives a mixture rich in styrene. Polyethylene and polypropylene, more than half of all plastics, lack a clean unzipping path at all: their radicals transfer hydrogen and cut chains at random, giving waxes and oils.

Poly(α-methylstyrene) shows the other edge of the knife: its ceiling is 63 °C, so it is useless as a hot-drink cup but easy to return to monomer. That is the design brief for new circular polymers — a ceiling above every temperature of use, below every temperature of recycling — met by polyesters from γ-butyrolactone (Hong and Chen) and, by a different route, by polydiketoenamines that come apart in acid (Christensen et al.).

Where things stand, late 2026. PET depolymerisation is commercial in several countries, and an engineered cutinase depolymerised 90 per cent of PET in about ten hours, though the first commercial-scale enzymatic plant has been slower to arrive than announced. Polyolefins remain the hard case: pyrolysis plants exist, but they are small against the 350 million tonnes of plastic waste a year, need much energy, and give broad product mixtures; "recycled content" claims made by mass-balance accounting are contested. Catalytic routes are elegant but at laboratory scale, such as tandem dehydrogenation and ethenolysis of polyethylene to propene. Most life-cycle studies rank mechanical recycling first where it works, chemical recycling ahead of incineration for mixed waste, and reduction and reuse ahead of both. For polymer structure itself, see Lumenary Materials' polymers and sustainable materials.

ExplanationH

Replacing PFAS

Environmental chemistry explained what PFAS are, why they persist and how drinking-water limits are set. The frontier question is different: what do you use instead, and what do you do with what is already there?

What fluorine does that nothing else does. A surface packed with CF₃ groups has a surface energy around 6 mN/m, lower than any hydrocarbon (about 20 to 30 mN/m for CH₃ and CH₂ surfaces). A liquid beads up only if the surface's energy is well below its own surface tension. Water, at 72.8 mN/m, is repelled by hydrocarbon waxes and silicones as well as by fluorocarbons; oils, at 25 to 35 mN/m, are repelled almost only by fluorinated surfaces. The Owens–Wendt model in the exercises puts numbers on it: water on a CH₃ surface about 113°, on CF₃ about 133°; hexadecane spreads on polyethylene, sits at about 38° on CH₃, and beads at about 94° only on CF₃. That is why fluorine-free textile finishes keep raincoats dry but lose stain and oil resistance, and why fluorine-free firefighting foams, now standard at many airports and fire services, needed new formulations rather than a drop-in swap.

Essential use and regrettable substitution. Replacing one PFAS with another has gone badly before: PFOA, used to make fluoropolymers, was replaced by shorter or ether-linked acids such as HFPO-DA ("GenX"), which proved more mobile in water and are now regulated in turn. The essential-use concept sorts uses into those that are non-essential (stain-resistant carpet), substitutable (many foams and coatings) and essential for now (some medical devices and semiconductor processes). In 2023 five European countries proposed restricting the whole class, some ten thousand substances, through the EU's chemicals agency; as of 2026 it is still being evaluated sector by sector. The energy transition has its own dilemma: PFSA membranes in PEM electrolysers and fuel cells, and fluoropolymer binders in lithium-ion batteries, are fluorinated, and hydrocarbon membranes and water-processable binders are active research.

Destroying PFAS, and recovering the fluorine. Water treatment concentrates PFAS on carbon or resins; the concentrate must then be destroyed, by high-temperature incineration, supercritical water oxidation or electrochemical oxidation. Two recent results point somewhere better. In 2022 William Dichtel's group showed that perfluorocarboxylic acids mineralise below about 120 °C once their carboxylate end is removed, because the chain then unzips through fluoride loss. In 2025 Véronique Gouverneur's group reported a mechanochemical method with potassium phosphate that breaks down many PFAS, including PTFE, and recovers the fluorine as salts that can make new fluorochemicals — closing a loop for fluorite, itself a critical raw material.

ExplanationH

Critical elements and recycling

A raw material is critical when it is economically important and its supply is at risk: few producing countries, little substitution, little recycling. The EU's Critical Raw Materials Act of 2024 and the United States' critical-minerals list both include the battery metals (lithium, cobalt, nickel, graphite), the platinum-group metals, gallium and germanium (placed under Chinese export licensing in 2023), and the rare earths, which met further export controls in 2025.

The rare earths show why chemistry is the bottleneck. Mining them is not unusual; separating them is. The lanthanoids differ mainly in the steady shrinking of their 3+ ions across the series (the lanthanoid contraction), so neighbours such as neodymium and praseodymium, both needed for the strongest permanent magnets, behave almost identically. Industry separates them by solvent extraction in cascades of mixer–settlers, each stage enriching one element by a separation factor β\beta as low as about 1.5 for such neighbours. Most of the world's separation capacity is in China. The number of stages follows the Fenske equation, Nmin⁡=ln⁡ ⁣[xtop/(1−xtop)xbottom/(1−xbottom)]/ln⁡βN_{\min} = \ln\!\left[\dfrac{x_{\text{top}}/(1-x_{\text{top}})}{x_{\text{bottom}}/(1-x_{\text{bottom}})}\right] / \ln\beta, which the exercises use; real plants run hundreds of stages to split a mixture of many lanthanoids. New ideas include proteins that bind lanthanoids with extraordinary selectivity, such as a lanmodulin dimer that separates neodymium from dysprosium, and electrochemical and membrane separations.

Recycling as urban mining. The world generated about 62 million tonnes of electronic waste in 2022, of which only about 22 per cent was documented as collected and recycled (Global E-waste Monitor). Circuit boards can be richer in gold than most ores, and platinum-group metals from car catalysts are recycled at high rates because they are concentrated and valuable. Most other elements in electronics are spread thin across alloys and composites, and the minimum-work argument returns: the more dilute and mixed the element, the more energy and chemistry recovery takes. Battery recycling, by smelting or by leaching and solvent extraction, is taken up in The chemistry of energy and in Lumenary Materials' battery materials.

Design for disassembly is the practical answer: an element that stays concentrated in an identifiable part (a magnet that can be unclipped, a catalyst that can be removed) is cheap to recover; the same element smeared through a shredded mixture is not. Separation costs are decided at the design stage.

ComputationC

Life-cycle assessment: a worked comparison

Catalysis and green chemistry outlined life-cycle assessment (LCA), standardised in ISO 14040 and 14044, in four phases: goal and scope (the functional unit, the service compared, and the system boundary); inventory (every input and emission inside the boundary); impact assessment (multiplying each emission by a characterisation factor, such as a global warming potential); and interpretation, including sensitivity to assumptions.

raw materialsextractionproductionuseend of liferecyclingenergy and materials in; emissions out at every stagecradle-to-gatecradle-to-grave
System boundaries in life-cycle assessment. A cradle-to-gate study stops at the factory gate; a cradle-to-grave study follows the product through use and disposal. Recycling sends material back into production. Energy and materials enter, and emissions leave, at every stage.

The worked comparison: greenhouse-gas emissions of one kilogram of hydrogen at the plant gate (functional unit; cradle-to-gate boundary) by three routes. Grey hydrogen comes from steam reforming of methane, with no capture. Blue adds CO₂ capture, which needs extra gas. Green comes from electrolysis. Two inventory items dominate: the CO₂ from the plant, and methane leaked along the gas supply chain, characterised with the IPCC's global warming potential for fossil methane, 29.8 over 100 years or 82.5 over 20. Every number below is an assumption stated in the code; change it and watch the ranking move.

python
# Cradle-to-gate greenhouse-gas intensity of 1 kg of hydrogen by three routes, in kg CO2e per kg H2.
# Every parameter is a stated assumption: change them and watch the ranking move.
GWP = {"GWP100": 29.8, "GWP20": 82.5}     # fossil methane, IPCC AR6

CO2_DIRECT = 9.0                  # kg CO2 per kg H2 from reforming plus burning gas for heat
CH4_USED = CO2_DIRECT * 16.04 / 44.01   # kg of methane consumed per kg H2 (all its carbon ends as CO2)
UPSTREAM_CO2 = 0.3                # kg CO2 per kg H2 from producing and moving the gas (excluding leaks)

def fossil(leak, capture=0.0, energy_penalty=0.0, gwp="GWP100"):
    """Grey (capture = 0) or blue hydrogen. leak = fraction of produced gas lost to the air."""
    gas = CH4_USED * (1 + energy_penalty)                # capture needs extra gas
    leaked = gas * leak / (1 - leak)
    direct = CO2_DIRECT * (1 + energy_penalty) * (1 - capture)
    return direct + UPSTREAM_CO2 * (1 + energy_penalty) + leaked * GWP[gwp]

def electrolytic(grid, kwh=52.0):
    """Hydrogen from electricity whose life-cycle intensity is `grid` kg CO2e per kWh."""
    return kwh * grid

rows = [
    ("grey, 1.5 % leak",                 fossil(0.015)),
    ("blue, 60 % capture, 1.5 % leak",   fossil(0.015, 0.60, 0.10)),
    ("blue, 90 % capture, 0.5 % leak",   fossil(0.005, 0.90, 0.10)),
    ("blue, 90 % capture, 1.5 % leak",   fossil(0.015, 0.90, 0.10)),
    ("blue, 90 % capture, 3.0 % leak",   fossil(0.030, 0.90, 0.10)),
    ("blue, 90 %, 1.5 % leak, GWP20",    fossil(0.015, 0.90, 0.10, "GWP20")),
    ("electrolysis, wind (0.011)",       electrolytic(0.011)),
    ("electrolysis, solar PV (0.048)",   electrolytic(0.048)),
    ("electrolysis, grid at 0.25",       electrolytic(0.25)),
    ("electrolysis, coal power (0.82)",  electrolytic(0.82)),
]
for name, value in rows:
    bar = "#" * min(round(value * 2), 40)
    print(f"{name:36s} {value:6.1f}  {bar}{'>' if value > 20 else ''}")

eu_limit = 0.30 * 94e-3 * 120          # 70 % below 94 g CO2e/MJ, times 120 MJ/kg (lower heating value)
print(f"\nEU low-carbon threshold: {eu_limit:.2f} kg CO2e/kg H2")
print(f"electrolysis beats grey below {fossil(0.015) / 52:.3f} kg/kWh,"
      f" meets the EU threshold below {eu_limit / 52:.3f} kg/kWh")
Runs entirely in your browser, in an isolated worker. Nothing is sent to a server.

The life-cycle intensities of wind, solar and coal electricity are the IPCC's 2014 median estimates; the grid at 0.25 kg/kWh is roughly a European average. Read the bars as an argument, not a verdict. Grey hydrogen sits near 11 kg CO₂e per kg. Blue hydrogen ranges from 1.9 to 5.9 depending on the capture rate, the leak rate and the time horizon, which is exactly why a 2021 paper concluded that blue hydrogen was worse than burning gas and a 2022 response found it could be much better: their assumptions differed on these very parameters. Electrolysis is the best route on wind or solar power and the worst on a coal grid; it beats grey hydrogen only below about 0.21 kg CO₂e per kWh and meets the EU's low-carbon threshold of 3.38 kg CO₂e per kg (70 per cent below a fossil comparator) only below about 0.065.

ExplanationH

The professional view

Who does this work. Process engineers design capture plants, electrolysers and recycling lines; LCA practitioners use databases such as ecoinvent and software such as openLCA; techno-economic analysts put costs beside emissions, because a technology nobody can afford reduces nothing. Chemists supply what all of them need: sorbents that survive humid flue gas, catalysts without iridium, polymers designed for depolymerisation, separations that work on dilute streams, and analytical methods precise enough to check claims.

How to read claims. Ask for the functional unit, the boundary and the assumptions; ask whether a number is a laboratory result, a pilot or an operating plant; ask what happens to the by-products; and check whether "capture" includes permanent storage. Tracking from the IEA and data from the IPCC are good places to start. In each area of this lesson, honest progress is real and measurable — and slower than announcements.

Optional Christian reflection · To dress it and to keep it

A reflection from the Christian tradition that inspires Lumenary Academy. It is not part of the mathematics and you are welcome to skip it.

In Genesis the first human work is a charge to tend a garden: "to dress it and to keep it" (Genesis 2:15, King James Version). Much of this lesson is the arithmetic of keeping — what un-making a mess costs, and how much cheaper it is not to make it. Stewardship in this sense is less a feeling than a habit of counting carefully and choosing well. This reflection is optional, and nothing in the lesson depends on it.

Common misconceptionsH

Common misconceptions

Misconception

"Carbon capture makes a power plant zero-emission." Plants are typically designed to capture about 90 per cent of their flue-gas CO₂ and often capture less in operation; the capture energy burns more fuel; and upstream methane leaks are untouched. Captured CO₂ that is used in a fuel returns to the air when the fuel is burned.

Misconception

"Hydrogen is a clean fuel." Hydrogen is an energy carrier, as clean as the way it was made. From coal it is among the highest-carbon fuels per unit of energy; from wind electricity it is among the lowest.

Misconception

"Chemical recycling can turn any plastic waste back into new plastic." Only polymers with an accessible ceiling temperature or a cleavable backbone (PMMA, PET, nylon-6 and new designs) return cleanly to monomer. Polyolefins mostly give oils that must be cracked again, with large energy and yield losses.

Misconception

"Fluorine-free means safe." Replacements must be assessed in their own right. Some PFAS replacements were themselves PFAS; some fluorine-free alternatives have other hazards or poorer performance that leads to more use. Substitution is chemistry and toxicology, not labelling.

Exercises

Exercises

1. Sustainable chemistry

Warm-upHuman

Using Wmin⁡=−RTln⁡xW_{\min} = -RT\ln x, what is the least work needed to extract a small amount of CO₂ as the pure gas from air at 25 °C (298.15 K), where its mole fraction is 420 ppm? Give the answer in kJ per mole, to the nearest whole number.

2. Sustainable chemistry

Warm-upHuman

Green ammonia is made by N₂ + 3 H₂ → 2 NH₃ with electrolytic hydrogen. How many kilograms of hydrogen does one tonne of ammonia contain, to the nearest kilogram? (H₂ 2.016 g/mol, NH₃ 17.031 g/mol.)

3. Sustainable chemistry

CoreHuman

A new polymer designed for chemical recycling has ΔHp=−42\Delta H_p = -42 kJ/mol and ΔSp∘=−120\Delta S_p^\circ = -120 J/(mol K) (monomer at 1 mol/L). What is its ceiling temperature at 1 mol/L of monomer, in °C, to the nearest degree?

4. Sustainable chemistry

CoreHuman

A 50:50 mixture of neodymium and praseodymium is to be split into a neodymium product of 99.5 % purity and a praseodymium product of 99.5 % purity, by solvent extraction with a separation factor β=1.5\beta = 1.5 per stage. Using the Fenske equation for the minimum number of ideal stages (at total reflux), how many whole stages are needed at least?

5. Sustainable chemistry

CoreComputational

In the Owens–Wendt model, a liquid with surface tension γl=γld+γlp\gamma_l = \gamma_l^d + \gamma_l^p (dispersive and polar parts) meets a solid with surface energy components γsd,γsp\gamma_s^d, \gamma_s^p at a contact angle θ\theta given by

cos⁡θ=−1+2(γsdγld+γspγlp)γl.\cos\theta = -1 + \frac{2\left(\sqrt{\gamma_s^d\gamma_l^d} + \sqrt{\gamma_s^p\gamma_l^p}\right)}{\gamma_l}.

Write contact_angle(gs_d, gs_p, gl_d, gl_p) returning θ\theta in degrees, 0.0 if the liquid spreads (cos⁡θ≥1\cos\theta \ge 1) and 180.0 if cos⁡θ≤−1\cos\theta \le -1. Then compare water (γd=21.8\gamma^d = 21.8, γp=51.0\gamma^p = 51.0 mN/m) and hexadecane (27.5, 0) on a CF₃ surface (6, 0), a CH₃ surface (22, 0) and polyethylene (31, 0).

Runs entirely in your browser, in an isolated worker. Nothing is sent to a server.

6. Sustainable chemistry

StretchComputational

Using the lesson's life-cycle model, write fossil(leak, capture=0.0, penalty=0.0, gwp=29.8), the kg CO₂e per kg of hydrogen from steam reforming, where leak is the fraction of produced gas lost to the air, capture the fraction of direct CO₂ captured, penalty the fractional extra gas burned for capture, and gwp the global warming potential of methane. Gas use, direct CO₂ and upstream CO₂ all scale with (1+penalty)(1 + \text{penalty}); leaked methane is gas used × leak/(1−leak)\times\ \text{leak}/(1-\text{leak}).

Then write max_leak(capture, penalty, target, gwp=29.8): the leak rate at which the intensity equals target, or 0.0 if the target cannot be met even with no leaks.

Runs entirely in your browser, in an isolated worker. Nothing is sent to a server.

7. Sustainable chemistry

CoreHuman

An outdoor-clothing maker replaces a fluorinated water-repellent finish with a silicone one. Customers report that jackets still shed rain but now stain with cooking oil. What is the best explanation?

Choose one

8. Sustainable chemistry

CoreHuman

A report states: "Producing a glass bottle emits 300 g CO₂e and producing a PET bottle 100 g CO₂e, so PET is three times better for the climate." What is the most important flaw?

Choose one

Further reading and resources